Cadenflow

BACKGROUND

Please describe your company and position.

I am the CEO of GUD Bags

Describe what your company does in a single sentence.

Backpacks

OPPORTUNITY / CHALLENGE

What specific goals or objectives did you hire Cadenflow to accomplish?

• Customer Service AI Automation for cost reduction and revenue focused optimisations.

SOLUTION

How did you find Cadenflow?

Referral

Why did you select Cadenflow over others?

Great culture fit

Describe the scope of work in detail. Please include a summary of key deliverables.

GUD Bags makes backpacks and everyday carry gear. We sell through two storefronts — one international, one domestic — in two languages, and we handle customer support ourselves. Support was spread across three places: warranty questions came through Instagram DMs, returns through email, and some customers just called. We had no helpdesk software, nothing linking a message to the order behind it, and we only answered during business hours. International customers writing in the evening waited until the next morning. We brought in Cadenflow to work out how to fix that and to start building it. Key deliverables: • An operational deep-dive into our support setup. They went through how every channel is actually handled day to day, where messages get lost, and which question types make up the bulk of the volume — pre-sale sizing and materials questions, order tracking, warranty eligibility, returns. • A written FAQ and knowledge base, built from our own past replies, covering pre-orders and production timelines, shipping and tracking, our warranty and its exclusions, returns and exchanges, and product-specific questions. Written in both Ukrainian and English. • A support tone-of-voice guide. This one mattered to us more than it might sound. It documented exactly how our support should read — greetings, grammar rules including correct vocative forms in Ukrainian, warmth without corporate stiffness, and language rules for our market. It's now the reference anyone joining our support can be handed on day one. • A CX strategy playbook. They reviewed around 20 customer support and AI vendors — from free open-source options to enterprise platforms — with current pricing, and told us plainly which tier fits a business our size and which ones to avoid. It included a three-stage framework (get everything into one inbox, write down our knowledge, then automate gradually), separate playbooks for email, Instagram DM and Telegram, a realistic monthly cost, a 90-day rollout plan we could execute ourselves, and a note on the EU AI disclosure rules that apply to us.

RESULTS & FEEDBACK

What were the measurable outcomes from the project that demonstrate progress or success?

Increased speed of response to potential buyer's questions

% of automated replies to general questions (WISMO, product questions)

Describe their project management. Did they deliver items on time? How did they respond to your needs?

We worked directly with the founder. One person, start to finish — no account manager in between, no briefing someone new each time. It started with a long structured discovery session. He came prepared, having already gone through both our stores and worked out how our support was routed publicly, so the questions were specific rather than generic. The next day we had a written one-page scope so both sides knew what was being built. After that we got regular updates as things progressed, and each deliverable arrived as a proper written document rather than notes in a chat. Everything landed when it was supposed to. When we pushed back — particularly on anything touching our brand voice, which we're protective of — it was taken seriously and reworked, not defended. He was clear about what he could and couldn't commit to, and when the engagement wound down he handed over the full playbook anyway so we could keep going without him. That was not something he had to do.

What was your primary form of communication with Cadenflow?

• In-Person Meeting
• Virtual Meeting

What did you find most impressive or unique about this company?

The honesty about AI. Everyone selling AI support tools quotes you automation rates of 70–85%. Cadenflow told us the real-world numbers are closer to 40–55%, and that for emotional or money-related conversations it's far lower — and that this is fine, because those conversations should stay human anyway. That kind of framing is not what you expect from someone whose business is building AI systems. He also talked us out of spending money. Several of the well-known AI support platforms we'd been considering start at several hundred dollars a month and are built for ticket volumes far above ours. He said so directly and pointed us at a setup costing a fraction of that. The bigger insight was that our main problem wasn't a lack of AI at all — it was that our channels were fragmented and our policies had never been written down anywhere a tool or a new hire could read them. Fixing that came first. Most vendors would have skipped straight to selling us the bot. And he treats support as a revenue channel rather than a cost to minimise, which changed how we think about it internally.

Are there any areas for improvement or something Cadenflow could have done differently?

The engagement ended before we had a system running live. We got the analysis, the knowledge base, the voice guide and a full strategy we can execute — but not a deployed, working setup, which is where we'd originally hoped to land. To be fair, we were told directly rather than left guessing, and everything was handed over in a usable state. Early on, a lot of questions were asked about numbers we simply didn't track — return rates, cost per return, response times by channel. It was the right thing to ask, but we'd have benefited from being told upfront which figures to prepare before the first call, so we weren't guessing on the spot. If we ran it again, we'd start smaller — pick our single busiest channel and get something modest working there first, rather than mapping the whole operation before shipping anything.

BACKGROUND

Please describe your company and position.

 I am the CEO of an ecommerce company 

Describe what your company does in a single sentence.

We sell healthcare products online.

OPPORTUNITY / CHALLENGE

What specific goals or objectives did you hire Cadenflow to accomplish?

• Customer Support AI Automation to reduce operational costs & save founder's time.

SOLUTION

How did you find Cadenflow?

Referral

Why did you select Cadenflow over others?

• Close to my geographic location
• Pricing fit our budget
• Great culture fit
• Good value for cost
• Referred to me

How many teammates from Cadenflow were assigned to this project?

1 Employee

Describe the scope of work in detail. Please include a summary of key deliverables.

We hired Cadenflow to build an AI-powered customer support system for our Shopify store. We sell a personal-care device direct-to-consumer across Europe and the US, and support was handled entirely by our small internal team through a shared Gmail inbox — no helpdesk, no support staff. As order volume grew, response times slipped and the workload was becoming unsustainable.

The engagement covered discovery, design, build, and a supervised production rollout.

Key deliverables:

• Support inbox audit. They analyzed roughly 400 of our historical support conversations, categorized every ticket by intent, and gave us a written breakdown of what our customers actually contact us about. This alone gave us visibility we'd never had.
• Automated ticket triage. A system that reads each incoming email, identifies the customer and their order in Shopify automatically, and classifies the request — order status, shipping/address issues, product questions, defect claims, refund and return requests, or cases needing a human.
• AI-drafted replies in our brand voice. They built a knowledge base from our return policy, defect-handling process, and product FAQs, plus a documented tone-of-voice guide. The system drafts replies in the customer's own language (we get English, German, Spanish, Italian, French).
• Human approval workflow. Every draft was routed to our team for one-click approve, edit, or reject before sending — with full order context attached. Nothing went out unreviewed.
• Safety guardrails. Sensitive cases — chargebacks, legal language, injury claims, disputed refunds — were flagged and routed to a human by design.
• Staged rollout plan. A defined path from review-only, to assisted replies, to selective automation, with accuracy thresholds required before advancing each stage.
• Ongoing production reviews. Weekly analysis of live system performance against real tickets, with documented findings and fixes shipped in response.

The system ran on infrastructure dedicated to us and hosted in the EU, with GDPR-conscious handling of customer data throughout.

RESULTS & FEEDBACK

What were the measurable outcomes from the project that demonstrate progress or success?

>60% of automated responses to common emails (WISMO, Refund, Change Address)

Describe their project management. Did they deliver items on time? How did they respond to your needs?

We worked directly with the founder, so there was no account-manager layer — the person writing the code was the person on our calls. For us that was the main advantage: questions got answered the same day, usually within a couple of hours.

The project was structured in clear milestones rather than one big delivery. Each stage had a defined scope and a written spec we reviewed before work started, so we always knew what was being built and what "done" meant. Deliverables landed on the dates we agreed.

Communication was a mix of scheduled calls and day-to-day messaging. During the live rollout we got weekly reviews of how the system was actually performing on our real tickets — including the cases it got wrong, not just the ones it got right. That level of honesty was unusual and made it easy to trust the reporting.

When we flagged issues — replies that didn't match our tone, messages the system routed to us that it should have handled, edge cases specific to our return policy — they were investigated and fixed in the next iteration. Nothing was dismissed or left hanging.

Overall: organized, transparent, and fast to respond.

What was your primary form of communication with Cadenflow?

• Virtual Meeting
• Email or Messaging App

What did you find most impressive or unique about this company?

Two things set them apart from the other AI vendors we spoke to.

First, the engineering rigor. Most people pitching AI support offered a chatbot you switch on and hope for the best. Cadenflow refused to automate anything until it had been measured against our own historical tickets, and built a testing setup that scored the system's accuracy before any stage was expanded. When they said a path was ready, they could show us the numbers behind it.

Second, the safety instinct. From day one the design assumed the AI would sometimes be wrong, and made sure that being wrong was harmless — every reply passed a human, and sensitive categories like disputes, legal language, and injury claims were deliberately routed to us and never handled automatically. For a business where one bad reply to an angry customer can turn into a chargeback, that mattered more than raw automation percentage.

They also took the time to actually learn our return policy and our tone, rather than dropping a generic tool on top of our inbox.

Are there any areas for improvement or something Cadenflow could have done differently?

The main one is the early ramp-up. In the first weeks of live use, the system was too cautious — it flagged a large share of messages for us to handle but often passed them over without a suggested reply attached, which meant our team was still writing responses from scratch. It also didn't initially recognize messages coming through our website's contact form, so those customers weren't being matched to their orders. Both were identified and fixed, but we'd have gotten value sooner if that behavior had been caught in testing rather than in production.

Related to that: the approval notifications we received sometimes lacked the customer and order details, which made it harder to review a draft quickly and confidently. A small thing, but it slowed us down early on.

If we did it again, we'd push for a longer dry-run against our real inbox before going live.